Restaurant Technology and AI

Website Chatbot vs Phone AI: Where Guests Ask

Which guest questions come in by website chat and which by phone, what each channel handles well at a restaurant, and how to keep one set of answers.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing8 min read

On this page
  1. Same guests, different moments
  2. Which questions arrive where
  3. What each channel does well, and where each fails
  4. Side by side
  5. One set of answers for both channels
  6. What the phone side is worth
  7. Where an SI Employee fits
  8. A two-week channel audit
  9. Frequently asked questions
  10. References
Restaurant Technology and AI: Website Chatbot vs Phone AI: Where Guests Ask49%: of reservations at 55 top US restaurants were made by phone, against 4% on the restaurant's websiteRESTAURANT TECHNOLOGY AND AIWebsite Chatbot vs Phone AI:Where Guests Ask49%of reservations at 55 top US restaurants were made by phone,against 4% on the restaurant's websiteAva for RestaurantsInsightsRestaurant Technology and AI: Website Chatbot vs Phone AI: Where Guests Ask49%: of reservations at 55 top US restaurants were made by phone, against 4% on the restaurant's websiteRESTAURANT TECHNOLOGY AND AIWebsite Chatbot vsPhone AI: WhereGuests Ask49%of reservations at 55 top USrestaurants were made by phone,against 4% on the restaurant's websiteAva for Restaurants
Figure source: Cornell University Center for Hospitality Research, 2012[1]

Key takeaways

  • Chat catches questions asked while reading the menu. The phone catches questions asked while driving, deciding or running late.
  • Bookings, changes, large parties and catering still lean on the phone[1], so cover the phone before polishing chat.
  • Chat helps guests who prefer typing, but it does not replace relay calls, which ADA-covered businesses must accept[2].
  • Write one set of approved answers and feed both channels from it, or chat, phone and Google will drift apart.

Same guests, different moments

Guests who type and guests who call are often the same people a few hours apart. At lunch someone reads your menu and types “is the pasta gluten-free?” into a chat box. At 5:40 the same person calls from a car because the table for six is running twenty minutes late. The website catches questions asked while reading. The phone catches questions asked while moving, deciding or worried. Choosing between a website chatbot and phone answering starts with knowing which of those moments costs you more when nobody responds.

For bookings, the phone has held its ground for a long time. When Cornell researchers collected reservation data from 55 top US restaurants, 49% of reservations had been made by telephone and 25% through a multi-restaurant booking site, while 4% came through the restaurant's own website; walk-ins made up the other 22% of business[1]. In the same research, about two-thirds of diners who had never booked online said they wanted to talk to someone to make a booking[1].

Where reservations and business came from at 55 top US restaurantsAverage shares reported to Cornell researchers, 2012. Walk-ins are a share of business rather than reservations.
Where reservations and business came from at 55 top US restaurantsDonut chart. Telephone: 49%; Multi-restaurant booking site: 25%; Walk-ins: 22%; Restaurant website: 4%.Telephone: 49%Multi-restaurant booking site: 25%Walk-ins: 22%Restaurant website: 4%2012
  • Telephone 49%
  • Multi-restaurant booking site 25%
  • Walk-ins 22%
  • Restaurant website 4%

Source: Cornell University Center for Hospitality Research, 2012[1]

Show the data
ItemValue
Telephone49%
Multi-restaurant booking site25%
Walk-ins22%
Restaurant website4%

Those figures are from 2012 and online booking has grown since, yet the shape still holds for the calls that matter most: changes, groups and anything with a condition attached. Your own logs are the better guide, and the audit at the end of this article takes two weeks.

Which questions arrive where

Where each kind of guest request tends to fit. Sort two weeks of your own chat logs and call logs by these rows before deciding anything.
Guest needWebsite chatPhone
Menu, dietary labels, pricesStrong: the guest is already on the menu pageCommon when the menu is hard to find on a phone screen
Hours, parking, directionsFineVery common from people already driving
Table for two tonightPoint to the booking linkRequest taken for the host to confirm, or booked by staff
Party of eight or moreCan collect details; many hosts prefer to talkStrong: seating, timing, high chairs, cake
Running late or changing a bookingRarely usedStrong: guests reach for the phone
Catering and private eventsInquiry form or chat leadStrong: budget, guest count and date questions
Allergy questionsLink the allergen guide, then hand offHand off to a manager
ComplaintsCollect details, hand offHand off to a manager

Changes and late arrivals are the clearest phone calls. In an earlier Cornell survey of 1,230 frequent diners, respondents believed guests should call when plans change, yet often found it hard to reach a restaurant to change a reservation[3]. A chat box on the website does nothing for a guest who is already on the highway and has never opened your site.

Example: a 90-seat bistro tags two weeks of contacts. Chat brings 140 conversations, and 100 of them are menu, hours and parking questions the website already answers somewhere. The phone brings 610 calls. About 240 are the same kind of quick question, but 150 are bookings or changes, 30 are parties of eight or more and 12 are catering or private-event inquiries, and the missed-call report shows 95 calls rang out, most between 6 and 8 pm. The chat log tells the owner the website needs a clearer menu page. The call log tells her where the money is leaking.

Run the same exercise before buying either tool. The rows that matter are the ones where an unanswered contact costs a table, an order or an event, and the hours when those contacts arrive.

What each channel does well, and where each fails

Chat earns its place with written answers to written questions. The guest is already reading, so a link to the allergen guide, the private-dining menu or the booking page does the job. It leaves a transcript, and it helps guests who would rather type than talk. Mobile ordering is now mainstream, with 57% of adults having recently used it[4], and 16% of adults are smartphone-only internet users with no home broadband[5], so the chat box has to work on a small screen.

Chat fails in two familiar ways. It only reaches guests who are already on your site, and it can trap them. The Consumer Financial Protection Bureau's report on chatbots describes “doom loops”: repetitive dead ends that happen when a customer's question falls outside what the bot can do, with no way through to a person[6]. A restaurant chat that answers “I can only help with menu questions” to a guest asking about a cancelled anniversary dinner is a doom loop with a dessert menu attached.

The phone earns its place with urgency and detail. A caller can explain a party of nine with two wheelchairs and a birthday cake in ninety seconds. The phone fails at peak, when the host stand is seating guests and the line rings out, and it fails after closing, when the questions keep coming and nobody is there. That is the gap phone answering software exists to close.

Accessibility cuts across both. Guests who are deaf or hard of hearing can reach you by phone through the free relay service at 7-1-1, and businesses covered by the ADA must accept relay calls and treat them like any other call[2]. A chat box is a welcome extra route, not a substitute, and it must itself be usable: the Justice Department's web guidance lists food and drink establishments among businesses covered by Title III and names inaccessible online forms as a barrier[7]. Serving guests who are deaf, blind or use wheelchairs covers the host-stand side.

Side by side

Website chat and phone answering for a restaurant
Website chatbotPhone answering
Who it reachesVisitors already on your websiteAnyone with your number, including from Maps and search
Best atMenus, hours, links and written detailBookings, changes, large parties and catering
Urgent questionsWait for a replyAnswered while the caller is on the line
Handoff to staffA message for laterTransfer to the manager path, then a backup
Record for the teamChat transcriptSummary by email, recording and transcript
Peak serviceUnaffected by a busy floorNeeds overflow cover when the host is seating

The right column assumes calls are forwarded to an answering system on overflow and after hours, not left to ring out.

If you can afford only one, cover the phone first when your call log shows unanswered calls during service, regular large-party bookings or catering inquiries. Put chat first when the phone is under control and your inbox is full of the same five menu questions; a strong FAQ page sometimes answers those before anyone needs a chatbot. The guide to AI chat on restaurant websites covers setting chat up well once you get there.

One set of answers for both channels

The real risk of running both is disagreement. The chat says the kitchen closes at 10, the phone says 9:30, Google says 10:30, and the guest who arrives at 9:45 believes whichever answer suits them. Guests also want to know what they are dealing with: 76% of Americans say it is extremely or very important to be able to tell whether text, pictures and video were made by AI or by people[8]. Label the chat as automated, and let the phone greeting say the same.

Example: changing holiday hours once, everywhere
  1. 01Your team

    Manager updates the sheet

    Thanksgiving hours go into the house information document first.

  2. 02Your team

    Website and chat updated

    The FAQ page and the chat tool's answers are changed from that text.

  3. 03Your team

    Phone answers updated

    The same wording goes into the phone assistant's approved information.

  4. 04Your team

    Listing special hours set

    Google's special hours match the sheet, word for word.

  5. 05Your team

    Both channels tested

    Ask the chat and call the line: both give the same closing time.

Example sequence. One owner, one document, every channel updated from it on the same day.

  • Write answers as full sentences a guest could read aloud, not shorthand like “kit cl 930”.
  • Give seasonal items and hours a start and end date, so nobody promises the patio in December.
  • Name one owner for the document and one backup.
  • Every Monday, ask the chat and the phone the same three questions and compare the answers.

Preparing restaurant information for AI tools has a template for the document. If messages are already scattered across calls, texts, DMs and email, one inbox for calls, texts, DMs and email shows how to give each channel an owner.

What the phone side is worth

A website chatbot cannot recover a call that rang out. To decide where the first dollar goes, price the phone calls you are losing now and compare that with the chat questions going unanswered.

ExampleBooking and order calls that ring out during service
Example: worked figures
InputExample
Calls to the restaurant per week250
Calls that ring out unanswered (%)18%
Calls that are new bookings or orders (%)45%
Share that would have booked or ordered (%)50%
Average check per booking or order$70
Unanswered calls a month195
From new customers88
Bookings that would have followed44
Monthly value at risk$3,069

How it is worked out: Calls a week × 4.33 weeks × unanswered share × new-customer share × booking rate × average value.

Assumptions

  • Phone is still a main booking channel: 49% of reservations at 55 top US restaurants were made by telephone, and about two-thirds of diners who had not booked online wanted to talk to someone.[1]
  • Example: 250 calls a week with 18% ringing out, mostly between 6 and 8 pm on Fridays and Saturdays.
  • Example: $70 is a table for two with drinks or a family takeout order. Use your own average check.

Example inputs. Your phone system's missed-call report and your POS average replace them.

Example: on these inputs about $3,070 a month in bookings and orders sits in calls nobody answered. Compare that with what an unanswered chat costs: usually a guest who reads the menu page instead, or calls. If the phone number is the larger one, the phone gets the budget first.

Where an SI Employee fits

An SI Employee covers the phone column. The restaurant forwards its number on overflow, after hours or all day; the SI Employee answers from house information you approve, takes reservation requests for your host to confirm, sends catering details to your catering manager, texts your official ordering link when the caller agrees (texting needs a registered SMS sender), and hands allergy questions and complaints to the manager you choose. Each call arrives as a summary by email, with the recording and transcript in the account. It does not run your website chat, so keep your chat tool reading the same approved document. More on AssistCall, and on the phone side call by call in how AI phone answering works for restaurants.

Fine dining shows the split clearly. In the fine-dining reservation line example, most tables book online, but the phone carries the hard calls: changes, large parties, allergies and private dinners.

A two-week channel audit

  1. Export both logs

    Fourteen days of chat transcripts, or website contact messages if you have no chat, and your phone system's call log with missed calls.

  2. Tag every contact

    Use the rows of the table above: menu, hours, small booking, large party, change, catering, allergy, complaint.

  3. Mark what went unanswered

    Calls that rang out and chats that sat without a reply, by hour of the day.

  4. Ask both channels the same questions

    Kitchen close tonight, holiday hours, private room capacity. Note every difference.

  5. Fix the document, then spend

    Correct the house information first. Then put the money on the channel with the most valuable unanswered contacts.

Frequently asked questions

Does a restaurant need both a website chatbot and phone AI?

Not necessarily. Start with whichever channel loses more valuable contacts. For most full-service restaurants that is the phone, because bookings, changes, large parties and catering still lean on it[1].

Can a website chatbot take reservations?

It can link to your booking page or collect a request. Unless it is connected to your reservation system, a person still confirms the table, the same as with a phone request.

Should the chatbot and the phone give identical answers?

The facts should be identical: hours, closures, policies, prices you choose to share. Feed both from one approved document and check them against each other every week.

Does website chat cover guests who are deaf or hard of hearing?

It helps, but it does not replace the phone. ADA-covered businesses must accept relay calls made through 7-1-1 and treat them like any other call[2], and the chat itself needs to be accessible[7].

Should we tell guests the chat or phone assistant is automated?

Yes. Most Americans say it matters to be able to tell whether content was made by AI or by people[8], and a plain label or greeting costs nothing.

References

  1. Cornell University Center for Hospitality Research (2012). The Role of Multi-Restaurant Reservation Sites in Restaurant Distribution Management (Kimes and Kies, Cornell Hospitality Report, 2012) (opens in a new tab) ↑

  2. US Department of Justice, Civil Rights Division (ADA.gov) (2026). ADA Requirements: Effective Communication (opens in a new tab) ↑

  3. Cornell University Center for Hospitality Research (Sheryl E. Kimes) (2008). A Consumer's View of Restaurant Reservations Policies (Cornell Hospitality Report, 2008) (opens in a new tab) ↑

  4. National Restaurant Association (2025). From Trend to Transformation: Off-Premises Dining Now Essential for Restaurant Consumers, Operators (opens in a new tab) ↑

  5. Pew Research Center (2026). Americans' Use of Mobile Devices in 2026 (Fact Sheet) (opens in a new tab) ↑

  6. Consumer Financial Protection Bureau (2023). Chatbots in consumer finance (opens in a new tab) ↑

  7. US Department of Justice, Civil Rights Division (ADA.gov) (2022). Guidance on Web Accessibility and the ADA (opens in a new tab) ↑

  8. Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑

Book a demo

See Ava handle your calls.

Book a short demo. We’ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo